Add stm32f4-0.14.0 benchmark

[Kobzol]

Time for compiling the whole crate did increase, but not by that much:

        stm32f4   stm32f4-0.14.0
check   13.13s     18.63s
debug   14.86s     19.37s
opt     14.13s     20.64s

What's quite interesting for this crate is that check takes almost as much time as debug/opt, so the main overhead is in rustc and not in LLVM here.

Instruction counts have increased, but seem reasonable (and much higher than for hello world):

(Don't open the Files changed tab. You have been warned. Better just check the second commit 😆 ).

[nnethercote]

Looks good, thank you. Please don't merge this until the current site breakage is resolved.

[rylev]

Running this particular benchmark took 10m20s making it the longest running benchmark we have (the second longest cargo-0.60.0 takes 6m21s). It seems like this cost might be worth it, but are we sure?

[nnethercote]

Good question. We run with "cargo_opts": "--features=stm32f405", does disabling that help at all?

[Kobzol]

The crate is basically empty if you don't use one of the features (actually it will not even build, build.rs will error out). At least one of the STM architectures has to be compiled.